An instrumented observability coverage method for system validation

In order to improve effectiveness and efficiency of post-silicon validation, we present a fault-symbol tracking method and a coverage metric that account for the limited observability in silicon and thus are useful for guiding validation test selection, test development, and design for debug. The coverage points targeted in this study are a set of fault-symbols, or ‘tags’, generated from each expression in a system model. Coverage is measured in simulation by tracking tags alongside dynamic information flows to user-defined or implicit observation points. Computation of the metric is performed based on high-level (C/C++) functional and behavioral models through compiler-inserted parallel fault-symbol tracking instrumentation, which offers high efficiency as well as compatibility with existing simulation flows. The coverage results from our initial implementation for a microcontroller instruction set simulator are compared with the statement and mutation coverages. The results show that the new coverage metric is more accurate than the statement coverage and can be computed in significantly shorter runtimes than the mutation coverage.